Drug overdose and alcohol related deaths
Introduction
Use of drugs and alcohol has substantial financial consequences. Every year, excessive alcohol use costs 249 billion dollars in lost productivity, medical expenditures, and criminal justice costs. In contrast, illicit drug usage costs 193 billion dollars in lost productivity, criminality, and medical costs.Communities can benefit from adopting and putting into practice policies that lessen prescription drug abuse and excessive alcohol consumption.
Source: https://www.countyhealthrankings.org/health-data/health-factors/health-behaviors/alcohol-and-drug-use
Our Question
Are there demographic and social factors that are predictors of drug overdose and alcohol-related incidents (e.g., driving accidents)?
Data
The data is collected by the University of Wisconsin Population Health Institute. It assigns a ranking based on health outcomes and health factors to each county in every state. There was originally 770 variables in the data set.Since the data set is so big, we decided that we should create another data set with only the variables we want to look at with relation to drug overdose deaths and alcohol impaired driving deaths. Predictor variables are drug overdose deaths and alcohol impaired driving deaths.
Here are some of the predictor variables we picked and the reason that we picked them:
Unemployment: Compared to the employed population, the unemployed population has worse health and higher death rates.
Median Income: The impact of median household income on the likelihood of poverty on individuals and families can have detrimental effects on their mental and physical well-being.
Disconnected Youth: Disconnected youth have a higher likelihood of experiencing violence, smoking, drinking, and using marijuana.
High School Graduation: A higher level of education is linked to a lower likelihood of smoking. Adults with higher levels of education typically have more employment opportunities and make more money overall.
Social Associations: Reduced participation in communal life and little social interaction with others are linked to higher rates of illness and early death. Compared to those with strong networks, those without strong social networks are less likely to choose healthy lifestyles.
(source: https://www.countyhealthrankings.org/health-data)
Upon inspection, we found that the variable of drug overdose deaths had lots of N/A values and we want to make sure this is clean as this is one of the main variables we want to look at. Also rename the data so that it can be read easier and look presentbale for any visuals.
Intial EDA
Looking at the states with respect to drug overdose deaths and alcohol impaired driving deaths. As a result, we can identify the areas with the most issues and, eventually, come up with solutions.
Just by observation, you can see that West Virginia has the highest amount of drug overdose deaths. And Montana has the most alcohol impaired driving deaths. By the predictors of both drug overdose deaths and alcohol impaired driving deaths, we can hopefully come up with some solutions and or regulations to potentially help these states with their ongoing problems.
Regression
So to eventually uncover the predictors, we need to conduct a regression analysis. This can assist us in deciding which variables should be taken into consideration and which ones not. But first we should find the best regression model for our data. This can help us get more accurate results.
Test and Train to find the best model for our data
First looking at drug overdose:
We should use a linear model for alcohol since there is no intersection with the other models
Now looking alcohol impaired deaths:
All of the models intersect so there isn’t a “best” model to work with it so we will use linear model since that is what we are using for alcohol
Linear Model
First lets look at alcohol impaired driving deaths
The substantial positive coefficients of variables such as “X65 and older,” “Unemployment,” “American Indian or Alaska native,” and “Median household income” indicate that these characteristics are linked to a rise in the number of alcohol-impaired driving deaths.
Variables such as “Hispanic” and “Female” have noteworthy negative coefficients, indicating that these attributes are linked to a reduction in deaths resulting from alcohol-impaired driving.
Now onto drug overdose deaths:
The significant positive coefficients of variables such as Unemployment, Female, and Children in single parent households indicate that these factors are linked to a rise in drug overdose deaths.
Social/Demographic affect health variables that them affect our response variables
Conclusion
So after looking at our data, we really cant conclude about any predictors that are demographic and social. What we think is that there are social and demographic factors that affect health factors that then are predictors of